 This is Stu Miniman with wikibond.org. Here we're SiliconANGLE TVs, live continuous coverage from VMworld 2012, here in Moscone North, the Hang Space, day three. Some of us are getting a little raspy in the voice from the long conversations we've had and pleased to be joining you for the networking panel. So networking is a huge discussion here. If you talk about the NICERA acquisition that just happened. The networking stack is a piece of the virtualization environment that everyone is focusing on. Storage gets a lot of play from us here, of course, as well as everybody there, but I think for the last year or so, everybody's saying that the next big innovation and changes in the marketplace are happening in the networking side. And as I said, pleased to have a nice distinguished panel to help walk through some of those changes. So joining me here, I'll walk through how is you from Big Switch, Greg Shearer from Broadcom, and Haseeb. Badani from Infanetta, sorry about that, Haseeb. So guys, welcome on theCUBE. Greg, you've been on before, but the other guys, try to bring to our community kind of the latest changes, the disruptions in the marketplace, helps extract the signal from the noise. Last year, networking was a big part. I remember right at VMworld, VXLan and NVGRE was the big discussion point. And here again, with the NICERA acquisition and Software Defined Data Center is going. So, Howie, I think the networking folks are going to know who you are, but can you give us a little bit of your background and start us off with what you see as the kind of Software Defined Data Center and network. So give us a minute or two on that. Thank you, Steve. Thank you, Steve, for inviting me to this place. This is my first time, that's why I'm a little bit nervous, but I'm the VP Engineering at Big Switch. Prior to Big Switch, I was at VMworld for almost a decade, and then running the entire networking strategy, vision, and engineering execution. I have been associated with server virtualization revolution from day one, and now with SDN journey from day one. It's such a blessing for me. And a lot of things VMworld innovated and then revolutionized over the last decade. It exposed a number of issues, and that's where SDN gets more important, because there are more and more boxes. There are more and more silos. And VMworld made a server-side provisioning management so flexible, automated, so much things. It exposed the slowness on the networking side. I often call the server side, it's the machine speed, and the networking side is the human speed. Now, it's such a disparity. It creates this opportunity for SDN to catch up. So that's why I'm excited at this SDN. Okay, great. Lots we're going to get into. So, Greg, when you talk about software really being a critical component, and the networking side, most of us that look at this say, merchant silicon is a real important piece of this. And of course, Broadcom supplies both the server side and the switch side chipset that many of the players, the majority of the players use. You guys made an announcement this week. Can you just give us an intro about what you're doing in this space and what's the dynamic in networking of hardware or software? You bet. So we're very excited. So we did, as you mentioned, our announcement. We made a couple of announcements, but we announced our Triton 2 switch. So it's a little like, we've been pregnant with Triton 2 for quite a while, and so it's wonderful to actually bring it to market and be able to talk about it openly. But it's the first device that has over 110 gig ports in a single piece of silicon. It also supports a very rich set of SDN manageable features. To be able to export that control plane, be managed through other frameworks. In addition to that, we also have a lot of offloads. So NVGRE and VXLAN are built into the silicon now, as opposed to just having to have specialized gateways, literally all of our XGS line moving forward will support this capability. So we're very excited about that. Inside the server, from a host perspective, we also have announced network virtualization acceleration from the standpoint of being able to take NVGRE and accelerate large send offload, large receive offload, but at a UDP basis, since NVGRE is really L2 over UDP. So great. So taking a lot of the protocols, baking it into the hardware, making it kind of transparent across all product lines. The control plane belongs in the host, but hardware does what hardware does best, and that's accelerate, that data plane. And we're very excited about that, but that's where we need the rest of the ecosystem and our partners to be able to, how we talked about the incredible volume of VMs and traffic that's going up, that management really needs partners to be able to manage the scale. It's just unprecedented. Okay, so, see, when I look at kind of the overall trend, yes, software defined data centers, interesting marketing term, that the long-term trend here is distributed systems. And of course, can you kind of just give us, want to understand, Infineta was founded for some of these changes in the way and distributed systems place. So, start us with kind of some of the big themes that you guys are seeing, give a little intro for yourself and your company. Sure, so I'm the chief product officer at Infineta, as you know. Infineta's claim to fame is that we are the only van optimization company focusing on data center to data center links. Now, the thing that's different about data center to data center links is, the links are larger, right? So, when we started the company some years ago, one of the things we had to think through was, how do we design a system that can sit on multiple trunks, you know, 10 gig trunks in a data center? I mean, this was pre-STN. So, the only thing that would work for us was Merchant Silicon. So, without Broadcom, you know, we don't have a play. So, we started a long time ago with the 56802, like, you know, an old strategy XGS, and the platform we're building right now uses a Trident Plus, and going forward we'll use Trident 2 as well. So, the thing that we look for, you know, when designing a system is, how, firstly, how do we take as much of the software work and push it to hardware, right? Just simplify it, accelerate it, push it out to hardware, so that we can focus on the work we do well, which is deduplication, de-septimization. Broadcom makes life very easy for us. But, going forward, you know, in terms of what we see, you know, two years ago, the big requirement our customers had was, well, you need to support VPLS in a large data center, or a large provider environment. If you do that stuff in software, we have a big problem, because it's going to be very slow. So, we have been relying on Broadcom to help us solve those kinds of problems. And going forward, some of our larger service provider type customers are asking us for, you know, NVGRE, VXLAN, in some cases, STT support. There, we are also planning on relying on Broadcom to help us solve that problem, because, I mean, today we'll do it in software, but really the right place to do this is in hardware, so we can scale to, you know, 100 gigabits or even higher going forward. So, Greg, I'm going to come back to you, because, you know, while 60% of applications are virtualized today, and VMware is dominant in this place, one of the things that we've heard very clearly is, it's not a homogenous environment. Multi-hypervisor is for real. Acquisitions like NYSERA, like dynamic ops, like Wanova, are all kind of VMware's thing. In many ways, the hypervisors commoditize, and there's going to be heterogeneous environments. You know, one of the things you guys have done, which was kind of news to me, is VXLAN and NVGRE. Usually there's the standards wars, and, you know, we have to get one, and we have to, you know, get the other. Can you fill us in as to, you know, what you've got coming down the pipe? Absolutely, and, you know, you're right. I mean, normally there's multiple camps, and we certainly see, especially in the virtualization world, there isn't going to be a single player. You know, certainly VMware is an incredibly dominant player in this market today, and will remain so for a long time, but, you know, you look at Zen and KVM hypervisors, Microsoft with, you know, the advent of Windows Server 2012, formerly known as Windows 8. You know, that's going to come on strong, and that's where, you know, NVGRE comes in, and so that the issue becomes, how do you tie together these, you know, dis-separate environments that are not going to adopt one homogenous, you know, environment, even when the controllers or the host software supports tunneling in the native format of, say, NVGRE or VXLAN, what do you do once it gets into the network? And the only way to handle that is through gateways, to go ahead and handle, you know, proxy, and be able to convert, you know, from one network into the other. The fortunate part is, is that NVGRE and VXLAN are very, very similar. You know, both have a 24-bit ID. You can think of it in a logical sense, as really having a 24-bit VLAN that just represents the tenant ID, you know, the VM at this point. You know, one tunnel's under, or over IP, the other over UDP, but, you know, the basic concepts are really ripe for, you know, the hardware to be able to migrate that tunnel from one to the other. But again, it's that control plane, you know, our partners to be able to figure out that gateway strategy. The hardware has all the capability and acceleration, but we really are looking forward to our partners helping us put together that rich ecosystem to be able to blend that environment. So it doesn't matter what environment, even, you know, some GRE-type environments that want maybe a little bit more than NVGRE, you know, NYSERA and others have talked about, you know, extending some of the capabilities that we find in NVGRE. So, very interesting. It sounds like, you know, come back a year from now, we'll really be talking about, you know, being able to stitch these multiple environments together. Absolutely. So, Howie, coming back to you, physical networking, virtual networking, you know, what's the balance of power these days? You know, where are things going? You know, what should, you know, if I'm not just a networking guy, but my CIO, how do I look at the network? Well, so from big switch perspective, we sort of look at it from three different perspectives. Number one, just like you said, stitching things together. We don't just decouple logical network from the physical network. That's one important thing to do, but we also stitching and the bridge, the virtual and the physical switch or network together. So that's sort of one way to look at it. The other way to look at it is, we stitch the protocols or technology from different vendors, different hypervisors, different switches, different VXLAN or NVGI protocols. So that's the sort of the multi-vendor heterogeneous environment. The number three thing is the openness. Openness is so important. People don't like vendor locking, right? So for us, we love open API, open standard, and then open source stuff. So, VXLAN and NVGI being the open standard, we love that. Okay, OpenStack Quantum? Yes, we actually have OpenStack Quantum Plugin as well. So I see, but we're getting low on the time for the panel here. When you talk about scalability of environments, how much does the WAN play into this? So the way I think about this is, eight years ago, seven, eight years ago, when I was at my first one of my position company, the problem we were solving was file access. And the biggest problem was, you give these application guys the ability to seamlessly go and pull data from a remote location, they'll go do it. They don't think about it, they go do it because the WAN is not their problem, somebody else's problem. I think with SDN, the same thing is happening. And now I can go get a VM or talk to a VM that is who knows where, it doesn't really matter, I don't care, right? Because that is happening because of Messera or Big Switch or whoever, now we're going to see more and more traffic on the WAN. So bread and butter for us was replication backup. These are well thought through processes, but more and more, I'm 100% sure this is what's coming. People are going to just blindly start pushing a lot of traffic because they can, and that's the good part, right? I can go do anything, I can do overlay tunnels, there's a gateways, whatever, right? And that's going to push traffic over the WAN, same as it did for tens of years, I mean at least since I've been in the industry, and that's their trend I see. Because of this ability for application developers to push traffic everywhere, they will, and because they will, you need more and more support for the WAN, the WAN needs to be stronger, the WAN needs to become the land effectively, right? And that's where VC have played for folks like us. But more importantly, I mean, the scales are not what they used to be. So now you've got to think about merchant silicon. Without this, you cannot solve it. Either you build your own ASIC, which is foolish, or use merchant silicon. There's no other ways to do it. That's what you've chosen, the merchant silicon path. All right. Greg? We're just a piggyback on that because boy, I completely agree with this. This is such an important concept because when you look at where we were several years ago, you know, we were at using vMotion, moving workloads from one server to the next within a rack. And then it became, well, we can kind of do that within multiple racks. The SDN and the new virtual networking paradigm allows us to do vMotion at literally a global scale. So whether or not your data center is in New York, San Francisco, or potentially in Iceland for power reasons, you can move your workloads around where you have capacity. And that is going to cause incredible load on your WAN environment. And the best part is, you know, even if the VMware says, hey, today you cannot do vMotion across five, you know, over five milliseconds or 10 milliseconds, you know what? A network guy may think about it, but an app guy, they don't care. Damn, they don't even know what that means. All right. So some of the long distance vMotion, really cool. Howie, I want to be able to give you the last word here. It sounds like you have something you want to say, but also can you speak to kind of automation and simplicity because, you know, as you said, kind of network has, how do we get beyond network as the human speed? I like that. And make it, you know, work. Well, to me, it's sort of the three steps journey. Number one is the VMware innovator so much, make the server side so automated, so flexible. Now the networking needs to step on and then being very, a lot more flexible. But the end game is to really have a virtual networking covering both the virtual and the physical altogether. You know, and I also covering not just the LEGO 2, not just the WAN optimization. It's the entire LEGO 227 needs to be flexible enough. The networking service needs to follow the workload immediately, anything, anywhere, anytime. That's the Nirvana. I think in the WISD SDN, we will get there. That's why I'm so excited about this. Okay, so gentlemen, I apologize we're so short on time here. Recommend that all of our viewers, you know, check out all of our panelists. There's definitely blogs and papers that you can find from all these gentlemen and presentations that their folks are doing here and demos and the like. So this is Stu Miniman with wikibond.org and we will be back with our next guest right after this brief break. Stay tuned.